In preparation for adding POSIX clockwait variants of timedwait functions,
add a clockid_t parameter to futex_abstimed_wait functions and pass
CLOCK_REALTIME from all callers for the time being.
Replace lll_futex_timed_wait_bitset with lll_futex_clock_wait_bitset
which takes a clockid_t parameter rather than the magic clockbit.
* sysdeps/nptl/lowlevellock-futex.h,
sysdeps/unix/sysv/linux/lowlevellock-futex.h: Replace
lll_futex_timed_wait_bitset with lll_futex_clock_wait_bitset that
takes a clockid rather than a special clockbit.
* sysdeps/nptl/lowlevellock-futex.h: Add
lll_futex_supported_clockid so that client functions can check
whether their clockid parameter is valid even if they don't
ultimately end up calling lll_futex_clock_wait_bitset.
* sysdeps/nptl/futex-internal.h,
sysdeps/unix/sysv/linux/futex-internal.h
(futex_abstimed_wait, futex_abstimed_wait_cancelable): Add
clockid_t parameter to indicate which clock the absolute time
passed should be measured against. Pass that clockid onto
lll_futex_clock_wait_bitset. Add invalid clock as reason for
returning -EINVAL.
* sysdeps/nptl/futex-internal.h,
sysdeps/unix/sysv/linux/futex-internal.h: Introduce
futex_abstimed_supported_clockid so that client functions can check
whether their clockid parameter is valid even if they don't
ultimately end up calling futex_abstimed_wait.
* nptl/pthread_cond_wait.c (__pthread_cond_wait_common): Remove
code to calculate relative timeout for
__PTHREAD_COND_CLOCK_MONOTONIC_MASK and just pass CLOCK_MONOTONIC
or CLOCK_REALTIME as required to futex_abstimed_wait_cancelable.
* nptl/pthread_rwlock_common (__pthread_rwlock_rdlock_full)
(__pthread_wrlock_full), nptl/sem_waitcommon (do_futex_wait): Pass
additional CLOCK_REALTIME to futex_abstimed_wait_cancelable.
* nptl/pthread_mutex_timedlock.c (__pthread_mutex_timedlock):
Switch to lll_futex_clock_wait_bitset and pass CLOCK_REALTIME
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
With commit f0b2132b35 ("ld.so:
Support moving versioned symbols between sonames [BZ #24741]"), the
dynamic linker will find the definition of vfork in libc and binds
a vfork reference to that symbol, even if the soname in the version
reference says that the symbol should be located in libpthread.
As a result, the forwarder (whether it's IFUNC-based or a duplicate
of the libc implementation) is no longer necessary.
On older architectures, a placeholder symbol is required, to make sure
that the GLIBC_2.1.2 symbol version does not go away, or is turned in
to a weak symbol definition by the link editor. (The symbol version
needs to preserved so that the symbol coverage check in
elf/dl-version.c does not fail for old binaries.)
mips32 is an outlier: It defined __vfork@@GLIBC_2.2, but the
baseline is GLIBC_2.0. Since there are other @@GLIBC_2.2 symbols,
the placeholder symbol is not needed there.
Since hppa is not an outlier anymore regarding LLL_LOCK_INITIALIZER value,
we can now assume it 0 for all architectures.
Checked on a build for all major ABIs.
* nptl/nptl-init.c (__pthread_initialize_minimal_internal): Remove
initialization for LLL_LOCK_INITIALIZER different than 0.
* nptl/old_pthread_cond_broadcast.c (__pthread_cond_broadcast_2_0):
Assume LLL_LOCK_INITIALIZER being 0.
* nptl/old_pthread_cond_signal.c (__pthread_cond_signal_2_0): Likewise.
* nptl/old_pthread_cond_timedwait.c (__pthread_cond_timedwait_2_0):
Likewise.
* nptl/old_pthread_cond_wait.c (__pthread_cond_wait_2_0): Likewise.
* sysdeps/nptl/libc-lockP.h (__libc_lock_define_initialized): Likewise.
This patch move the single-thread syscall optimization defintions from
syscall-cancel.h to new header file single-thread.h and also move the
cancellation definitions from pthreadP.h to syscall-cancel.h.
The idea is just simplify the inclusion of both syscall-cancel.h and
single-thread.h (without the requirement of including all pthreadP.h
defintions).
No semantic changes expected, checked on a build for all major ABIs.
* nptl/pthreadP.h (CANCEL_ASYNC, CANCEL_RESET, LIBC_CANCEL_ASYNC,
LIBC_CANCEL_RESET, __libc_enable_asynccancel,
__libc_disable_asynccancel, __librt_enable_asynccancel,
__libc_disable_asynccancel, __librt_enable_asynccancel,
__librt_disable_asynccancel): Move to ...
* sysdeps/unix/sysv/linux/sysdep-cancel.h: ... here.
(SINGLE_THREAD_P, RTLD_SINGLE_THREAD_P): Move to ...
* sysdeps/unix/sysv/linux/single-thread.h: ... here.
* sysdeps/generic/single-thread.h: New file.
* sysdeps/unix/sysdep.h: Include single-thread.h.
* sysdeps/unix/sysv/linux/futex-internal.h: Include sysdep-cancel.h.
* sysdeps/unix/sysv/linux/lowlevellock-futex.h: Likewise.
Checked on x86_64-linux-gnu and i686-linux-gnu.
* nptl/tst-rwlock6.c: Use libsupport. This also happens to fix a
small bug where only tv.tv_usec was checked which could cause an
erroneous pass if pthread_rwlock_timedrdlock incorrectly took more
than a second.
* nptl/tst-rwlock7.c, nptl/tst-rwlock9.c, nptl/tst-rwlock14.c: Use
libsupport.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Checked on x86_64-linux-gnu and i686-linux-gnu.
* nptl/tst-sem5.c(do_test): Use xclock_gettime, timespec_add and
TEST_TIMESPEC_NOW_OR_AFTER from libsupport.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Don't run nptl/tst-eintr1 by normal make check because it can spuriously
break testing on various linux kernels. (Currently this affects the
aarch64 glibc buildbot machine which regularly fails and loses test
results.)
[BZ #24537]
* nptl/Makefile: Move tst-eintr1 to xtests.
Checked on x86_64-linux-gnu and i686-linux-gnu.
* nptl/tst-sem5.c: Remove unused headers. Add <support/check.h>.
(do_test) Use libsupport test macros rather than hand-coded
conditionals and error messages. Ensure that sem_init returns zero
rather than not -1. Use <support/test-driver.c> rather than
test-skeleton.c.
* nptl/tst-sem13.c: Add <support/check.h>. (do_test) Use libsupport
test macros rather than hand-coded conditionals and error messages.
Use <support/test-driver.c> rather than test-skeleton.c.
This patch refactor how hp-timing is used on loader code for statistics
report. The HP_TIMING_AVAIL and HP_SMALL_TIMING_AVAIL are removed and
HP_TIMING_INLINE is used instead to check for hp-timing avaliability.
For alpha, which only defines HP_SMALL_TIMING_AVAIL, the HP_TIMING_INLINE
is set iff for IS_IN(rtld).
Checked on aarch64-linux-gnu, x86_64-linux-gnu, and i686-linux-gnu. I also
checked the builds for all afected ABIs.
* benchtests/bench-timing.h: Replace HP_TIMING_AVAIL with
HP_TIMING_INLINE.
* nptl/descr.h: Likewise.
* elf/rtld.c (RLTD_TIMING_DECLARE, RTLD_TIMING_NOW, RTLD_TIMING_DIFF,
RTLD_TIMING_ACCUM_NT, RTLD_TIMING_SET): Define.
(dl_start_final_info, _dl_start_final, dl_main, print_statistics):
Abstract hp-timing usage with RTLD_* macros.
* sysdeps/alpha/hp-timing.h (HP_TIMING_INLINE): Define iff IS_IN(rtld).
(HP_TIMING_AVAIL, HP_SMALL_TIMING_AVAIL): Remove.
* sysdeps/generic/hp-timing.h (HP_TIMING_AVAIL, HP_SMALL_TIMING_AVAIL,
HP_TIMING_NONAVAIL): Likewise.
* sysdeps/ia64/hp-timing.h (HP_TIMING_AVAIL, HP_SMALL_TIMING_AVAIL):
Likewise.
* sysdeps/powerpc/powerpc32/power4/hp-timing.h (HP_TIMING_AVAIL,
HP_SMALL_TIMING_AVAIL): Likewise.
* sysdeps/powerpc/powerpc64/hp-timing.h (HP_TIMING_AVAIL,
HP_SMALL_TIMING_AVAIL): Likewise.
* sysdeps/sparc/sparc32/sparcv9/hp-timing.h (HP_TIMING_AVAIL,
HP_SMALL_TIMING_AVAIL): Likewise.
* sysdeps/sparc/sparc64/hp-timing.h (HP_TIMING_AVAIL,
HP_SMALL_TIMING_AVAIL): Likewise.
* sysdeps/x86/hp-timing.h (HP_TIMING_AVAIL, HP_SMALL_TIMING_AVAIL):
Likewise.
* sysdeps/generic/hp-timing-common.h: Update comment with
HP_TIMING_AVAIL removal.
This patch removes CLOCK_THREAD_CPUTIME_ID and CLOCK_PROCESS_CPUTIME_ID support
from clock_gettime and clock_settime generic implementation. For Linux, kernel
already provides supports through the syscall and Hurd HTL lacks
__pthread_clock_gettime and __pthread_clock_settime internal implementation.
As described in clock_gettime man-page [1] on 'Historical note for SMP
system', implementing CLOCK_{THREAD,PROCESS}_CPUTIME_ID with timer registers
is error-prone and susceptible to timing and accurary issues that the libc
can not deal without kernel support.
This allows removes unused code which, however, still incur in some runtime
overhead in thread creation (the struct pthread cpuclock_offset
initialization).
If hurd eventually wants to support them it should either either implement as
a kernel facility (or something related due its architecture) or in system
specific implementation.
Checked on aarch64-linux-gnu, x86_64-linux-gnu, and i686-linux-gnu. I also
checked on a i686-gnu build.
* nptl/Makefile (libpthread-routines): Remove pthread_clock_gettime and
pthread_clock_settime.
* nptl/pthreadP.h (__find_thread_by_id): Remove prototype.
* elf/dl-support.c [!HP_TIMING_NOAVAIL] (_dl_cpuclock_offset): Remove.
(_dl_non_dynamic_init): Remove _dl_cpuclock_offset setting.
* elf/rtld.c (_dl_start_final): Likewise.
* nptl/allocatestack.c (__find_thread_by_id): Remove function.
* sysdeps/generic/ldsodefs.h [!HP_TIMING_NOAVAIL] (_dl_cpuclock_offset):
Remove.
* sysdeps/mach/hurd/dl-sysdep.c [!HP_TIMING_NOAVAIL]
(_dl_cpuclock_offset): Remove.
* nptl/descr.h (struct pthread): Rename cpuclock_offset to
cpuclock_offset_ununsed.
* nptl/nptl-init.c (__pthread_initialize_minimal_internal): Remove
cpuclock_offset set.
* nptl/pthread_create.c (START_THREAD_DEFN): Likewise.
* sysdeps/nptl/fork.c (__libc_fork): Likewise.
* nptl/pthread_clock_gettime.c: Remove file.
* nptl/pthread_clock_settime.c: Likewise.
* sysdeps/unix/clock_gettime.c (hp_timing_gettime): Remove function.
[HP_TIMING_AVAIL] (realtime_gettime): Remove CLOCK_THREAD_CPUTIME_ID
and CLOCK_PROCESS_CPUTIME_ID support.
* sysdeps/unix/clock_settime.c (hp_timing_gettime): Likewise.
[HP_TIMING_AVAIL] (realtime_gettime): Likewise.
* sysdeps/posix/clock_getres.c (hp_timing_getres): Likewise.
[HP_TIMING_AVAIL] (__clock_getres): Likewise.
* sysdeps/unix/clock_nanosleep.c (CPUCLOCK_P, INVALID_CLOCK_P):
Likewise.
(__clock_nanosleep): Remove CPUCLOCK_P and INVALID_CLOCK_P usage.
[1] http://man7.org/linux/man-pages/man2/clock_gettime.2.html
This patch assumes realtime clock support for nptl and thus removes
all the associated code.
For __pthread_mutex_timedlock the fallback usage for the case where
lll_futex_timed_wait_bitset it not set define is also removed. The
generic lowlevellock-futex.h always define it, so for NPTL code the
check always yield true.
Checked on x86_64-linux-gnu and i686-linux-gnu.
* nptl/nptl-init.c (__have_futex_clock_realtime,
__have_futex_clock_realtime): Remove definition.
(__pthread_initialize_minimal_internal): Remove FUTEX_CLOCK_REALTIME
check test for !__ASSUME_FUTEX_CLOCK_REALTIME.
* nptl/pthread_mutex_timedlock.c (__pthread_mutex_timedlock): Assume
__ASSUME_FUTEX_CLOCK_REALTIME support.
* sysdeps/unix/sysv/linux/i386/lowlevellock.S: Likewise.
* sysdeps/unix/sysv/linux/x86_64/lowlevellock.S: Likewise.
* sysdeps/unix/sysv/linux/kernel-features.h
(__ASSUME_FUTEX_CLOCK_REALTIME): Remove.
* sysdeps/nptl/lowlevellock-futex.h (lll_futex_timed_wait_bitset):
Adjust comment.
In pthread_tryjoin if pd->tid == 0 then we will not block on a
futex operation because we will immediately see the join is already
complete and return. The comment is fixed to reflect that.
Signed-off-by: Carlos O'Donell <carlos@redhat.com>
sys/procfs.h was already using this sysdeps directory.
This avoids the need for nptl-specific wrapper headers under
include/, a generic location in the source tree.
After commit f1ac745583 ("arm: Use "nr"
constraint for Systemtap probes [BZ #24164]"), we load pd->result into
a register in the probe below:
/* Free the TCB. */
__free_tcb (pd);
}
else
pd->joinid = NULL;
LIBC_PROBE (pthread_join_ret, 3, threadid, result, pd->result);
However, at this point, the thread descriptor has been freed. If the
thread stack does not fit into the thread stack cache, the memory will
have been unmapped, and the program will crash in the probe.
Patch ce7eb0e903 ("nptl: Cleanup cancellation macros") changed the
join sequence for internal common __pthread_timedjoin_ex to use the
new macro lll_wait_tid. The idea was this macro would issue the
cancellable futex operation depending whether the timeout is used or
not. However if a timeout is used, __lll_timedwait_tid is called and
it is not a cancellable entrypoint.
This patch fixes it by simplifying the code in various ways:
- Instead of adding the cancellation handling on __lll_timedwait_tid,
it moves the generic implementation to pthread_join_common.c (called
now timedwait_tid with some fixes to use the correct type for pid).
- The llvm_wait_tid macro is removed, along with its replication on
x86_64, i686, and sparc arch-specific lowlevellock.h.
- sparc32 __lll_timedwait_tid is also removed, since the code is similar
to generic one.
- x86_64 and i386 provides arch-specific __lll_timedwait_tid which is
also removed since they are similar in functionality to generic C code
and there is no indication it is better than compiler generated code.
New tests, tst-join8 and tst-join9, are provided to check if
pthread_timedjoin_np acts as a cancellation point.
Checked on x86_64-linux-gnu, i686-linux-gnu, sparcv9-linux-gnu, and
aarch64-linux-gnu.
[BZ #24215]
* nptl/Makefile (lpthread-routines): Remove lll_timedwait_tid.
(tests): Add tst-join8 tst-join9.
* nptl/lll_timedwait_tid.c: Remove file.
* sysdeps/sparc/sparc32/lll_timedwait_tid.c: Likewise.
* sysdeps/unix/sysv/linux/i386/lll_timedwait_tid.c: Likewise.
* sysdeps/sysv/linux/x86_64/lll_timedwait_tid.c: Likewise.
* nptl/pthread_join_common.c (timedwait_tid): New function.
(__pthread_timedjoin_ex): Act as cancellation entrypoint is block
is set.
* nptl/tst-join5.c (thread_join): New function.
(tf1, tf2, do_test): Use libsupport and add pthread_timedjoin_np
check.
* nptl/tst-join8.c: New file.
* nptl/tst-join9.c: Likewise.
* sysdeps/nptl/lowlevellock-futex.h (lll_futex_wait_cancel,
lll_futex_timed_wait_cancel): Add generic macros.
* sysdeps/nptl/lowlevellock.h (__lll_timedwait_tid, lll_wait_tid):
Remove definitions.
* sysdeps/unix/sysv/linux/i386/lowlevellock.h: Likewise.
* sysdeps/unix/sysv/linux/sparc/lowlevellock.h: Likewise.
* sysdeps/unix/sysv/linux/x86_64/lowlevellock.h: Likewise.
* sysdeps/sparc/sparc32/lowlevellock.c (__lll_timedwait_tid):
Remove function.
* sysdeps/unix/sysv/linux/i386/lowlevellock.S (__lll_timedwait_tid):
Likewise.
* sysdeps/unix/sysv/linux/x86_64/lowlevellock.S: Likewise.
* sysdeps/unix/sysv/linux/lowlevellock-futex.h
(lll_futex_timed_wait_cancel): New macro.
The clone.S patch fixes 2 elfutils testsuite unwind failures, where the
backtrace gets stuck repeating __thread_start until we hit the backtrace
limit. This was confirmed by building and installing a patched glibc and
then building elfutils and running its testsuite.
Unfortunately, the testcase isn't working as expected and I don't know why.
The testcase passes even when my clone.S patch is not installed. The testcase
looks logically similarly to the elfutils testcases that are failing. Maybe
there is a subtle difference in how the glibc unwinding works versus the
elfutils unwinding? I don't have good gdb pthread support yet, so I haven't
found a way to debug this. Anyways, I don't know if the testcase is useful or
not. If the testcase isn't useful then maybe the clone.S patch is OK without
a testcase?
Jim
[BZ #24040]
* elf/Makefile (CFLAGS-tst-unwind-main.c): Add -DUSE_PTHREADS=0.
* elf/tst-unwind-main.c: If USE_PTHEADS, include pthread.h and error.h
(func): New.
(main): If USE_PTHREADS, call pthread_create to run func. Otherwise
call func directly.
* nptl/Makefile (tests): Add tst-unwind-thread.
(CFLAGS-tst-unwind-thread.c): Define.
* nptl/tst-unwind-thread.c: New file.
* sysdeps/unix/sysv/linux/riscv/clone.S (__thread_start): Mark ra
as undefined.
Commit 27761a1042 ("Refactor atfork
handlers") introduced a lock, atfork_lock, around fork handler list
accesses. It turns out that this lock occasionally results in
self-deadlocks in malloc/tst-mallocfork2:
(gdb) bt
#0 __lll_lock_wait_private ()
at ../sysdeps/unix/sysv/linux/x86_64/lowlevellock.S:63
#1 0x00007f160c6f927a in __run_fork_handlers (who=(unknown: 209394016),
who@entry=atfork_run_prepare) at register-atfork.c:116
#2 0x00007f160c6b7897 in __libc_fork () at ../sysdeps/nptl/fork.c:58
#3 0x00000000004027d6 in sigusr1_handler (signo=<optimized out>)
at tst-mallocfork2.c:80
#4 sigusr1_handler (signo=<optimized out>) at tst-mallocfork2.c:64
#5 <signal handler called>
#6 0x00007f160c6f92e4 in __run_fork_handlers (who=who@entry=atfork_run_parent)
at register-atfork.c:136
#7 0x00007f160c6b79a2 in __libc_fork () at ../sysdeps/nptl/fork.c:152
#8 0x0000000000402567 in do_test () at tst-mallocfork2.c:156
#9 0x0000000000402dd2 in support_test_main (argc=1, argv=0x7ffc81ef1ab0,
config=config@entry=0x7ffc81ef1970) at support_test_main.c:350
#10 0x0000000000402362 in main (argc=<optimized out>, argv=<optimized out>)
at ../support/test-driver.c:168
If no locking happens in the single-threaded case (where fork is
expected to be async-signal-safe), this deadlock is avoided.
(pthread_atfork is not required to be async-signal-safe, so a fork
call from a signal handler interrupting pthread_atfork is not
a problem.)
While debugging a kernel warning, Thomas Gleixner, Sebastian Sewior and
Heiko Carstens found a bug in pthread_mutex_trylock due to misordered
instructions:
140: a5 1b 00 01 oill %r1,1
144: e5 48 a0 f0 00 00 mvghi 240(%r10),0 <--- THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
14a: e3 10 a0 e0 00 24 stg %r1,224(%r10) <--- last THREAD_SETMEM of ENQUEUE_MUTEX_PI
vs (with compiler barriers):
140: a5 1b 00 01 oill %r1,1
144: e3 10 a0 e0 00 24 stg %r1,224(%r10)
14a: e5 48 a0 f0 00 00 mvghi 240(%r10),0
Please have a look at the discussion:
"Re: WARN_ON_ONCE(!new_owner) within wake_futex_pi() triggerede"
(https://lore.kernel.org/lkml/20190202112006.GB3381@osiris/)
This patch is introducing the same compiler barriers and comments
for pthread_mutex_trylock as introduced for pthread_mutex_lock and
pthread_mutex_timedlock by commit 8f9450a0b7
"Add compiler barriers around modifications of the robust mutex list."
ChangeLog:
[BZ #24180]
* nptl/pthread_mutex_trylock.c (__pthread_mutex_trylock):
One group of warnings seen with -Wextra is warnings for static or
inline not at the start of a declaration (-Wold-style-declaration).
This patch fixes various such cases for inline, ensuring it comes at
the start of the declaration (after any static). A common case of the
fix is "static inline <type> __always_inline"; the definition of
__always_inline starts with __inline, so the natural change is to
"static __always_inline <type>". Other cases of the warning may be
harder to fix (one pattern is a function definition that gets
rewritten to be static by an including file, "#define funcname static
wrapped_funcname" or similar), but it seems worth fixing these cases
with inline anyway.
Tested for x86_64.
* elf/dl-load.h (_dl_postprocess_loadcmd): Use __always_inline
before return type, without separate inline.
* elf/dl-tunables.c (maybe_enable_malloc_check): Likewise.
* elf/dl-tunables.h (tunable_is_name): Likewise.
* malloc/malloc.c (do_set_trim_threshold): Likewise.
(do_set_top_pad): Likewise.
(do_set_mmap_threshold): Likewise.
(do_set_mmaps_max): Likewise.
(do_set_mallopt_check): Likewise.
(do_set_perturb_byte): Likewise.
(do_set_arena_test): Likewise.
(do_set_arena_max): Likewise.
(do_set_tcache_max): Likewise.
(do_set_tcache_count): Likewise.
(do_set_tcache_unsorted_limit): Likewise.
* nis/nis_subr.c (count_dots): Likewise.
* nptl/allocatestack.c (advise_stack_range): Likewise.
* sysdeps/ieee754/dbl-64/s_sin.c (do_cos): Likewise.
(do_sin): Likewise.
(reduce_sincos): Likewise.
(do_sincos): Likewise.
* sysdeps/unix/sysv/linux/x86/elision-conf.c
(do_set_elision_enable): Likewise.
(TUNABLE_CALLBACK_FNDECL): Likewise.
The alignment of TLS variables is wrong if accessed from within a thread
for architectures with tls variant TLS_TCB_AT_TP.
For the main thread the static tls data is properly aligned.
For other threads the alignment depends on the alignment of the thread
pointer as the static tls data is located relative to this pointer.
This patch adds this alignment for TLS_TCB_AT_TP variants in the same way
as it is already done for TLS_DTV_AT_TP. The thread pointer is also already
properly aligned if the user provides its own stack for the new thread.
This patch extends the testcase nptl/tst-tls1.c in order to check the
alignment of the tls variables and it adds a pthread_create invocation
with a user provided stack.
The test itself is migrated from test-skeleton.c to test-driver.c
and the missing support functions xpthread_attr_setstack and xposix_memalign
are added.
ChangeLog:
[BZ #23403]
* nptl/allocatestack.c (allocate_stack): Align pointer pd for
TLS_TCB_AT_TP tls variant.
* nptl/tst-tls1.c: Migrate to support/test-driver.c.
Add alignment checks.
* support/Makefile (libsupport-routines): Add xposix_memalign and
xpthread_setstack.
* support/support.h: Add xposix_memalign.
* support/xthread.h: Add xpthread_attr_setstack.
* support/xposix_memalign.c: New File.
* support/xpthread_attr_setstack.c: Likewise.
For a full analysis of both the pthread_rwlock_tryrdlock() stall
and the pthread_rwlock_trywrlock() stall see:
https://sourceware.org/bugzilla/show_bug.cgi?id=23844#c14
In the pthread_rwlock_trydlock() function we fail to inspect for
PTHREAD_RWLOCK_FUTEX_USED in __wrphase_futex and wake the waiting
readers.
In the pthread_rwlock_trywrlock() function we write 1 to
__wrphase_futex and loose the setting of the PTHREAD_RWLOCK_FUTEX_USED
bit, again failing to wake waiting readers during unlock.
The fix in the case of pthread_rwlock_trydlock() is to check for
PTHREAD_RWLOCK_FUTEX_USED and wake the readers.
The fix in the case of pthread_rwlock_trywrlock() is to only write
1 to __wrphase_futex if we installed the write phase, since all other
readers would be spinning waiting for this step.
We add two new tests, one exercises the stall for
pthread_rwlock_trywrlock() which is easy to exercise, and one exercises
the stall for pthread_rwlock_trydlock() which is harder to exercise.
The pthread_rwlock_trywrlock() test fails consistently without the fix,
and passes after. The pthread_rwlock_tryrdlock() test fails roughly
5-10% of the time without the fix, and passes all the time after.
Signed-off-by: Carlos O'Donell <carlos@redhat.com>
Signed-off-by: Torvald Riegel <triegel@redhat.com>
Signed-off-by: Rik Prohaska <prohaska7@gmail.com>
Co-authored-by: Torvald Riegel <triegel@redhat.com>
Co-authored-by: Rik Prohaska <prohaska7@gmail.com>
This patch wraps all uses of *_{enable,disable}_asynccancel and
and *_CANCEL_{ASYNC,RESET} in either already provided macros
(lll_futex_timed_wait_cancel) or creates new ones if the
functionality is not provided (SYSCALL_CANCEL_NCS, lll_futex_wait_cancel,
and lll_futex_timed_wait_cancel).
Also for some generic implementations, the direct call of the macros
are removed since the underlying symbols are suppose to provide
cancellation support.
This is a priliminary patch intended to simplify the work required
for BZ#12683 fix. It is a refactor change, no semantic changes are
expected.
Checked on x86_64-linux-gnu and i686-linux-gnu.
* nptl/pthread_join_common.c (__pthread_timedjoin_ex): Use
lll_wait_tid with timeout.
* nptl/sem_wait.c (__old_sem_wait): Use lll_futex_wait_cancel.
* sysdeps/nptl/aio_misc.h (AIO_MISC_WAIT): Use
futex_reltimed_wait_cancelable for cancelabla mode.
* sysdeps/nptl/gai_misc.h (GAI_MISC_WAIT): Likewise.
* sysdeps/posix/open64.c (__libc_open64): Do not call cancelation
macros.
* sysdeps/posix/sigwait.c (__sigwait): Likewise.
* sysdeps/posix/waitid.c (__sigwait): Likewise.
* sysdeps/unix/sysdep.h (__SYSCALL_CANCEL_CALL,
SYSCALL_CANCEL_NCS): New macro.
* sysdeps/nptl/lowlevellock.h (lll_wait_tid): Add timeout argument.
(lll_timedwait_tid): Remove macro.
* sysdeps/unix/sysv/linux/i386/lowlevellock.h (lll_wait_tid):
Likewise.
(lll_timedwait_tid): Likewise.
* sysdeps/unix/sysv/linux/sparc/lowlevellock.h (lll_wait_tid):
Likewise.
(lll_timedwait_tid): Likewise.
* sysdeps/unix/sysv/linux/x86_64/lowlevellock.h (lll_wait_tid):
Likewise.
(lll_timedwait_tid): Likewise.
* sysdeps/unix/sysv/linux/clock_nanosleep.c (__clock_nanosleep):
Use INTERNAL_SYSCALL_CANCEL.
* sysdeps/unix/sysv/linux/futex-internal.h
(futex_reltimed_wait_cancelable): Use LIBC_CANCEL_{ASYNC,RESET}
instead of __pthread_{enable,disable}_asynccancel.
* sysdeps/unix/sysv/linux/lowlevellock-futex.h
(lll_futex_wait_cancel): New macro.
With upcoming fix for BZ#12683, pthread cancellation does not act for:
1. If syscall is blocked but with some side effects already having
taken place (e.g. a partial read or write).
2. After the syscall has returned.
The main change is due the fact programs need to act in syscalls with
side-effects (for instance, to avoid leak of allocated resources or
handle partial read/write).
This patch changes the NPTL testcase that assumes the old behavior and
also changes the tst-backtrace{5,6} to ignore the cancellable wrappers.
Checked on i686-linux-gnu, x86_64-linux-gnu, x86_64-linux-gnux32,
aarch64-linux-gnu, arm-linux-gnueabihf, powerpc64le-linux-gnu,
powerpc-linux-gnu, sparcv9-linux-gnu, and sparc64-linux-gnu.
* debug/tst-backtrace5.c (handle_signal): Avoid cancellable wrappers
in backtrace analysis.
* nptl/tst-cancel4.c (tf_write): Handle cancelled syscall with
side-effects.
(tf_send): Likewise.
In the read lock function (__pthread_rwlock_rdlock_full) there was a
code path which would fail to reload __readers while waiting for
PTHREAD_RWLOCK_RWAITING to change. This failure to reload __readers
into a local value meant that various conditionals used the old value
of __readers and with only two threads left it could result in an
indefinite stall of one of the readers (waiting for PTHREAD_RWLOCK_RWAITING
to go to zero, but it never would).
This patch eliminates the gen-py-const.awk variant of gen-as-const,
switching to use of gnu-as-const.py (with a new --python option) to
process .pysym files (i.e., to generate nptl_lock_constants.py), as
the syntax of those files is identical to that of .sym files.
Note that the generated nptl_lock_constants.py is *not* identical to
the version generated by the awk script. Apart from the trivial
changes (comment referencing the new script, and output being sorted),
the constant FUTEX_WAITERS, PTHREAD_MUTEXATTR_FLAG_BITS,
PTHREAD_MUTEXATTR_FLAG_PSHARED and PTHREAD_MUTEX_PRIO_CEILING_MASK are
now output as positive rather than negative constants (on x86_64
anyway; maybe not necessarily on 32-bit systems):
< FUTEX_WAITERS = -2147483648
---
> FUTEX_WAITERS = 2147483648
< PTHREAD_MUTEXATTR_FLAG_BITS = -251662336
< PTHREAD_MUTEXATTR_FLAG_PSHARED = -2147483648
---
> PTHREAD_MUTEXATTR_FLAG_BITS = 4043304960
> PTHREAD_MUTEXATTR_FLAG_PSHARED = 2147483648
< PTHREAD_MUTEX_PRIO_CEILING_MASK = -524288
---
> PTHREAD_MUTEX_PRIO_CEILING_MASK = 4294443008
This is because gen-as-const has a cast of the constant value to long
int, which gen-py-const lacks.
I think the positive values are more logically correct, since the
constants in question are in fact unsigned in C. But to reliably
produce gen-as-const.py output for constants that always (in C and
Python) reflects the signedness of values with the high bit of "long
int" set would mean more complicated logic needs to be used in
computing values.
The more correct positive values by themselves produce a failure of
nptl/test-mutexattr-printers, because masking with
~PTHREAD_MUTEXATTR_FLAG_BITS & ~PTHREAD_MUTEX_NO_ELISION_NP now leaves
a bit -1 << 32 in the Python value, resulting in a KeyError exception.
To avoid that, places masking with ~ of one of the constants in
question are changed to mask with 0xffffffff as well (this reflects
how ~ in Python applies to an infinite-precision integer whereas ~ in
C does not do any promotions beyond the width of int).
Tested for x86_64.
* scripts/gen-as-const.py (main): Handle --python option.
* scripts/gen-py-const.awk: Remove.
* Makerules (py-const-script): Use gen-as-const.py.
($(py-const)): Likewise.
* nptl/nptl-printers.py (MutexPrinter.read_status_no_robust): Mask
with 0xffffffff together with ~(PTHREAD_MUTEX_PRIO_CEILING_MASK).
(MutexAttributesPrinter.read_values): Mask with 0xffffffff
together with ~PTHREAD_MUTEXATTR_FLAG_BITS and
~PTHREAD_MUTEX_NO_ELISION_NP.
* manual/README.pretty-printers: Update reference to
gen-py-const.awk.
This patch does not have any functionality change, we only provide a spin
count tunes for pthread adaptive spin mutex. The tunable
glibc.pthread.mutex_spin_count tunes can be used by system administrator to
squeeze system performance according to different hardware capabilities and
workload characteristics.
The maximum value of spin count is limited to 32767 to avoid the overflow
of mutex->__data.__spins variable with the possible type of short in
pthread_mutex_lock ().
The default value of spin count is set to 100 with the reference to the
previous number of times of spinning via trylock. This value would be
architecture-specific and can be tuned with kinds of benchmarks to fit most
cases in future.
I would extend my appreciation sincerely to H.J.Lu for his help to refine
this patch series.
* manual/tunables.texi (POSIX Thread Tunables): New node.
* nptl/Makefile (libpthread-routines): Add pthread_mutex_conf.
* nptl/nptl-init.c: Include pthread_mutex_conf.h
(__pthread_initialize_minimal_internal) [HAVE_TUNABLES]: Call
__pthread_tunables_init.
* nptl/pthreadP.h (MAX_ADAPTIVE_COUNT): Remove.
(max_adaptive_count): Define.
* nptl/pthread_mutex_conf.c: New file.
* nptl/pthread_mutex_conf.h: New file.
* sysdeps/generic/adaptive_spin_count.h: New file.
* sysdeps/nptl/dl-tunables.list: New file.
* nptl/pthread_mutex_lock.c (__pthread_mutex_lock): Use
max_adaptive_count () not MAX_ADAPTIVE_COUNT.
* nptl/pthread_mutex_timedlock.c (__pthrad_mutex_timedlock):
Likewise.
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Signed-off-by: Kemi.wang <kemi.wang@intel.com>
There is a data-dependency between the fields of struct l_reloc_result
and the field used as the initialization guard. Users of the guard
expect writes to the structure to be observable when they also observe
the guard initialized. The solution for this problem is to use an acquire
and release load and store to ensure previous writes to the structure are
observable if the guard is initialized.
The previous implementation used DL_FIXUP_VALUE_ADDR (l_reloc_result->addr)
as the initialization guard, making it impossible for some architectures
to load and store it atomically, i.e. hppa and ia64, due to its larger size.
This commit adds an unsigned int to l_reloc_result to be used as the new
initialization guard of the struct, making it possible to load and store
it atomically in all architectures. The fix ensures that the values
observed in l_reloc_result are consistent and do not lead to crashes.
The algorithm is documented in the code in elf/dl-runtime.c
(_dl_profile_fixup). Not all data races have been eliminated.
Tested with build-many-glibcs and on powerpc, powerpc64, and powerpc64le.
[BZ #23690]
* elf/dl-runtime.c (_dl_profile_fixup): Guarantee memory
modification order when accessing reloc_result->addr.
* include/link.h (reloc_result): Add field init.
* nptl/Makefile (tests): Add tst-audit-threads.
(modules-names): Add tst-audit-threads-mod1 and
tst-audit-threads-mod2.
Add rules to build tst-audit-threads.
* nptl/tst-audit-threads-mod1.c: New file.
* nptl/tst-audit-threads-mod2.c: Likewise.
* nptl/tst-audit-threads.c: Likewise.
* nptl/tst-audit-threads.h: Likewise.
Signed-off-by: Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
All the required code already existed, and some of it was already
running.
AT_SYSINFO_EHDR is processed if NEED_DL_SYSINFO_DSO is defined, but it
looks like it always is. The call to setup_vdso is also unconditional,
so all that was left to do was setup the function pointers and use
them. This patch just deletes some #ifdef to enable that.
[BZ #19767]
* nptl/Makefile (tests-static): Add tst-cond11-static.
(tests): Likewise.
* nptl/tst-cond11-static.c: New File.
* sysdeps/unix/sysv/linux/Makefile (tests-static): Add
tst-affinity-static.
(tests): Likewise.
* sysdeps/unix/sysv/linux/sysdep-vdso.h: Check USE_VSYSCALL
instead of SHARED.
* sysdeps/unix/sysv/linux/sysdep.h (ALWAYS_USE_VSYSCALL): New.
(USE_VSYSCALL): Likewise.
* sysdeps/unix/sysv/linux/tst-affinity-static.c: New file.
* sysdeps/unix/sysv/linux/x86/libc-vdso.h: Check USE_VSYSCALL
instead of SHARED.
* sysdeps/unix/sysv/linux/x86_64/init-first.c: Don't check
SHARED.
* sysdeps/unix/sysv/linux/x86_64/sysdep.h (ALWAYS_USE_VSYSCALL):
New.
The race leads either to pthread_mutex_destroy returning EBUSY
or triggering an assertion (See description in bugzilla).
This patch is fixing the race by ensuring that the elision path is
used in all cases if elision is enabled by the GLIBC_TUNABLES framework.
The __kind variable in struct __pthread_mutex_s is accessed concurrently.
Therefore we are now using the atomic macros.
The new testcase tst-mutex10 is triggering the race on s390x and intel.
Presumably also on power, but I don't have access to a power machine
with lock-elision. At least the code for power is the same as on the other
two architectures.
ChangeLog:
[BZ #23275]
* nptl/tst-mutex10.c: New File.
* nptl/Makefile (tests): Add tst-mutex10.
(tst-mutex10-ENV): New variable.
* sysdeps/unix/sysv/linux/s390/force-elision.h: (FORCE_ELISION):
Ensure that elision path is used if elision is available.
* sysdeps/unix/sysv/linux/powerpc/force-elision.h (FORCE_ELISION):
Likewise.
* sysdeps/unix/sysv/linux/x86/force-elision.h: (FORCE_ELISION):
Likewise.
* nptl/pthreadP.h (PTHREAD_MUTEX_TYPE, PTHREAD_MUTEX_TYPE_ELISION)
(PTHREAD_MUTEX_PSHARED): Use atomic_load_relaxed.
* nptl/pthread_mutex_consistent.c (pthread_mutex_consistent): Likewise.
* nptl/pthread_mutex_getprioceiling.c (pthread_mutex_getprioceiling):
Likewise.
* nptl/pthread_mutex_lock.c (__pthread_mutex_lock_full)
(__pthread_mutex_cond_lock_adjust): Likewise.
* nptl/pthread_mutex_setprioceiling.c (pthread_mutex_setprioceiling):
Likewise.
* nptl/pthread_mutex_timedlock.c (__pthread_mutex_timedlock): Likewise.
* nptl/pthread_mutex_trylock.c (__pthread_mutex_trylock): Likewise.
* nptl/pthread_mutex_unlock.c (__pthread_mutex_unlock_full): Likewise.
* sysdeps/nptl/bits/thread-shared-types.h (struct __pthread_mutex_s):
Add comments.
* nptl/pthread_mutex_destroy.c (__pthread_mutex_destroy):
Use atomic_load_relaxed and atomic_store_relaxed.
* nptl/pthread_mutex_init.c (__pthread_mutex_init):
Use atomic_store_relaxed.
Set installed NPTL header as the expected one (instead of an
internal one for glibc testsuite) and add a hurd specific
stdc-predef with __STDC_NO_THREADS__.
Checked on both i686-linux-gnu and i686-gnu that both threads.h
and stdc-predef.h are the expected ones.
* nptl/threads.h: Move to ...
* sysdeps/nptl/threads.h: ... here.
* sysdeps/hurd/stdc-predef.h: New file.
